提出了应用光谱和纹理特征的高光谱成像技术早期检测番茄叶片早疫病的方法。 利用高光谱图像采集系统获取380～1 030 nm范围内71个染病和88个健康番茄叶片的高光谱图像, 同时采用主成分分析法(PCA)对高光谱图像进行处理。 选取染病和健康叶片感兴趣区域(region of interest, ROI)的光谱反射率值, 同时分别从前8个主成分的每幅主成分图像的ROI中提取对比度(Contrast)、 相关性(Correlation)、 熵(Entropy)和同质性(Homogeneity)4个灰度共生矩阵的纹理特征值, 再通过PCA和连续投影算法(SPA)结合最小二乘支持向量机(LS-SVM)构建番茄叶片早疫病的早期鉴别模型。 建立的6个模型中, 采用光谱反射率值的LS-SVM模型对番茄叶片早疫病的识别率最高, 达到100%。 结果表明, 应用高光谱成像技术检测番茄叶片早疫病是可行的。

Early detection of early blight on tomato leaves using hyperspectral imaging technique based on spectroscopy and texture was researched in the present study. Hyperspectral images of seventy-one infected and eighty-eight healthy tomato samples were captured by hyperspectral imaging system over the wavelength region of 380～1 030 nm and then were dimensioned by principal component analysis (PCA). Diffuse spectral response of region of interest (ROI) from hyperspectral image was extracted by ENVI software. At the same time, four features variables were extracted by texture analysis based on gray level co-occurrence matrix (GLCM) from each PC image of the first eight PCs including contrast, correlation, entropy and homogeneity, respectively. Then PCA and successive projections algorithm (SPA) were used to build least squares-support vector machine (LS-SVM) model to detect early blight on tomato leaves. Among the six models, LS-SVM model based on spectroscopy performed best with the discrimination of 100%. It was demonstrated that it is feasible to detect early blight on tomato leaves by hyperspectral imaging technique.